https://gcc.gnu.org/bugzilla/show_bug.cgi?id=115865
--- Comment #5 from Marek Polacek <mpolacek at gcc dot gnu.org> --- If the GIMPLE_TRY_FINALLY doesn't always have to be present, albeit empty, then I suppose the best fix would be to get rid of the empty finally block so that we don't have to check in many spots that it's not empty.